Re: AccrInt

["James R. Van Zandt" <jrv vanzandt mv com>]

Phillip J Shelton <shelton11 dingoblue net au> writes:

> "John S. Dey" wrote:
> > Its been brought up before on the gnumeric mailing list, that if, in
> > fact, errors are detected in Excel, what should the gnumeric
> > implementation do?  This is an issue that probably has already been
> > addressed so for what its worth I would implement the error in the
> > function (provided it can be replicated) and then add a feature that
> > would provide the correct calculation.
>
> I don't intend to un-Excel clone gnumeric.  If the spreadsheet functions
> need errors coded in then that is what has to happen. 

I would much prefer that both compatible and correct versions be
implemented.  There should be a over-all option to select which set is
used (the same way that setting the environment variable
POSIXLY_CORRECT changes the behavior of several other GNU programs).
If convenient, there could also be individual options to select
versions of individual features.

I can offer two reasons: First, some of us are more interested in
correctness than Excel compatibility.  Second, we can assume that
Microsoft application writers have some interest in correctness (at
least, more interest than their colleagues working on operating
systems :-).  Eventually, they are likely to issue a new version of
Excel with the correct implementation.  At that point, Gnumeric will
need the other implementation for compatibility as well as
correctness.

            - Jim Van Zandt


p.s. A run-time option would be nice.  However, a configuration-time
option would be better than nothing.